A Classified Warning Method for Heavy Overload in Distribution Networks Considering the Characteristics of Unbalanced Datasets

In order to achieve heavy overload warning and capacity planning for the distribution network, it is necessary classify of network. A network with classification method based on imbalanced dataset feature extraction proposed. Screening indicator set related overload, constructing a hierarchical prediction framework load situation, combining information such as power points, road construction, municipal planning, form line renovation plans. Based K-means clustering, undersampling used extract features from unbalanced classification, using decision trees basic learning unit. It includes multiple trained by Bagging integrated theory random subspace method. The forest algorithm realize detection weight grading, grading early realized according results. Tests have shown that this has good accuracy in predicting electrical loads can effectively solve problem excess caused light or no load, improving ability